Voice Activity Detection Using Combined Judgment Results

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing voice activity detection (VAD) technologies face inaccuracies in detecting voice activity, particularly in non-stationary noise and music signals, leading to inefficient processing and noticeable quality declines.

Innovation Solution

A voice activity detection method that combines the number of continuous active frames, average total SNR of all sub-bands, tonal signal flag, and at least two VAD judgment results to achieve a comprehensive judgment, improving accuracy by considering various parameters.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If traditional VAD algorithms are used, then the processing efficiency is improved by skipping inactive phases, but the detection accuracy deteriorates in non-stationary noise and music signals

Engineering Contradiction:
Improveprocessing efficiencyVSAvoiddetection accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent combines multiple VAD judgment results through a logical OR operation to make the final voice activity decision. This merging of multiple detection outcomes improves accuracy in complex acoustic environments while still enabling efficiency gains by identifying true inactive phases for skipping processing.

Inventive Principle:
Principle #5Merging (Combining)

2Use of energy by moving object

If VAD is applied to reduce processing in inactive phases, then energy consumption is reduced, but detection errors increase in music and non-stationary noise

Engineering Contradiction:
Improveenergy consumptionVSAvoiddetection reliability
Core Design Contradiction:
Use of energy by moving objectVSReliability

Solution Approach 1:

The system uses feedback from multiple VAD judgment results to improve detection reliability. By aggregating outcomes from different detection passes, the system achieves more reliable voice activity identification in challenging signals like music and non-stationary noise, reducing false positives that would waste energy on unnecessary processing.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3040991B1Voice activation detection method and device
Publication Date: 2021.04.14 ZTE CORP
  • EP3040991B1 patent drawingFigure 1
  • EP3040991B1 patent drawingFigure 2~3
  • EP3040991B1 patent drawingFigure 4

AI summary

Disclosed are a voice activity detection method and device, wherein the method includes: obtaining an ultimate combined VAD judgment result according to the number of continuous active frames, an average total SNR of all sub-bands, a tonal signal flag, and at least two existing VAD judgment results (107). The method and device are suitable for voice service, solve the problem of inaccurate VAD detection and realize highly accurate VAD judgment.